One of the major challenges in olefins cracking furnace operation is how to obtain key furnace measurements – for example, product yields and metal temperatures – reliably and in real time.
PSE’s CrackingMonitor is an innovative ‘virtual multisensor’ that uses your plant data in conjunction with a high-fidelity model of the cracking process to provide accurate, real-time values for key operational variables, as well as coke build-up over time along the furnace coil.
This can improve cracker economics by tens of millions of dollars per year.
How CrackingMonitor works …
CrackingMonitor utilises PSE’s unique Dynamic State Estimation technology coupled with a high-fidelity mathematical model to reconcile plant data and model predictions.
This allows it to calculate a consistent, accurate set of real-time operational values, as well as accurately monitor coke build-up over time along the coil length.
By providing better and more timely information on quantities such product yields and tube metal temperatures, CrackingMonitor enables much tighter furnace control than normally possible.
Increases in ethylene yield alone can improve plant economics by $1m to $3m per month on a large-scale plant – meaning that the payback time can be measured in weeks. Other benefits arise from:
- the ability to perform online margin optimisation on a furnace-by-furnace basis
- better furnace utilisation through advanced end-or-run (EOR) projection for decoking scheduling
- use of the calibrated furnace model in dynamic optimisation of the entire furnace section.
The overall benefits are estimated at $25m to $50m per year for a large ethylene plant.
The results shown below compare predictions with data. They were taken from a period of operation on a large ethylene plant in Saudi Arabia processing mainly ethane and propane feedstocks. Reasonably reliable ethylene yield measurements were available and could therefore be used to provide verification of model predictions.
Simulated yield prediction
The results below show simulated predictions (current practice) assuming fixed coking parameters, taking into account actual feed flowrate, controls, etc. against plant data (black line).
The grey line shows the ethylene yield. As the run progresses the predictions diverge significantly from the plant measurements.
The same is true for the Tube Metal Temperature, with the simulated value far from reality.
Accurate yield prediction
By contrast, the plots below shows the prediction using CrackingMonitor, where dynamic state estimation techniques ensure that the predictions follows the plant measurements closely.
The red line shows CrackingMonitor prediction following the plant measurement closely.
In the CrackingMonitor case, the predicted metal temperature homes in on the observed values over the course of the run.
Accurate monitoring of coking
The plot on the right shows coke build-up along the length of the coil over the run length.
Unlike traditional simulation-based approaches – which use a fixed coking kinetic model over the duration of the production run – CrackingMonitor uses plant data to continually update the cracking kinetic model based on observed plant data. This means that the coking model is effectively self-calibrating, and thus monitors the degree of coking very closely over the run.
Coke accumulation along the length of the coil over time